Positron AI Hits Unicorn Status With $230M Series B for Inference Hardware
The startup aims to challenge Nvidia's dominance by prioritizing energy efficiency and massive memory capacity for AI inference.
AI inference hardware startup Positron AI has raised $230 million in an oversubscribed Series B funding round, propelling the company to a post-money valuation exceeding $1 billion. The investment marks a significant bet on specialized silicon designed to reduce the power and memory bottlenecks currently plaguing large-scale AI deployments.
The funding round was co-led by ARENA Private Wealth, Jump Trading, and Unless, with strategic participation from Arm and the Qatar Investment Authority (QIA). Positron intends to use the capital to scale its energy-efficient hardware and accelerate the development of its next-generation silicon, dubbed "Asimov."
A Memory-First Approach
While the current AI chip market is dominated by Nvidia's compute-centric GPUs, Positron is positioning itself as a "memory-first" platform. This strategy targets the specific demands of long-context large language models (LLMs) and agentic workflows, which require massive amounts of data to be accessible quickly and efficiently.
The upcoming Asimov chip is designed to address these bottlenecks head-on. According to Positron AI CEO Mitesh Agrawal, the next-generation chip will deliver five times more tokens per watt in core workloads compared to Nvidia’s upcoming Rubin GPU. To support this efficiency, the Asimov chip is designed to ship with over 2,304 GB of RAM per device—a massive leap over the 384 GB expected for the Rubin architecture.
Positron is already shipping its first product, Atlas, a silicon system fabricated by Intel in the United States. The performance of the current hardware has already drawn praise from industry partners. Alex Davies, CTO of Jump Trading, noted that in their testing, Positron Atlas delivered roughly three times lower end-to-end latency than a comparable H100-based system on the inference workloads they evaluated.
Shifting the Hardware Race
This investment signals a broader shift in the AI hardware race, moving the focus from raw training power toward "inference efficiency." As the industry moves from developing models to deploying them at scale, the cost of electricity and the physical limits of memory bandwidth have become primary obstacles for enterprises.
By targeting these specific inefficiencies, Positron aims to lower the environmental impact and operational costs of deploying massive AI models. If successful, this approach could break Nvidia's current stranglehold on the AI infrastructure layer by offering a more sustainable and scalable alternative for the inference phase of the AI lifecycle.
What's Next
Positron is now focused on moving the Asimov chip into production, which is expected in early 2027. The industry will be watching closely to see if the startup can translate its theoretical efficiency gains and massive memory specs into real-world performance that can displace established GPU clusters in the data center.